LSTM-CNN网络在同步电机励磁绕组匝间短路故障预警中的应用  

Application of LSTM-CNN Network in Fault Waring of Inter-turn Short Circuit in Field Windings of Synchronous Generator

作　　者：李俊卿[1] 陈雅婷 LI Junqing;CHEN Yating(School of Electrical and Electronic EngineeringꎬNorth China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学电气与电子工程学院,河北保定071003

出　　处：《华北电力大学学报（自然科学版）》2020年第4期61-70,共10页Journal of North China Electric Power University：Natural Science Edition

摘　　要：人工智能与电力系统的结合日渐紧密,深度学习在实现电网智能化中起到了重要的作用。为了实现隐极同步电机励磁绕组匝间短路早期故障的在线预警,提出一种基于LSTM-CNN的神经网络预测模型。模型以同步电机正常运行时多个可测量物理量和励磁电流为网络输入、输出,利用历史数据进行网络训练,并根据拟合输入量与输出量之间的函数关系确定故障预警阈值。以相同的实验数据训练相同层数的LSTM网络、CNN网络与LSTM-CNN网络,结果证明LSTM-CNN网络在训练速度和拟合精度上的综合表现最佳。As deep learning has been playing an important role in realizing intelligent grid,this paper proposed an online fault warning model for field winding inter-turn short circuit of non-salient pole synchronous generator based on LSTM-CNN.The model took measurable physical quantities and excitation current under normal operation as network input and output,conducted network training with historical data,and determined fault warning threshold according to the function relation between fitting input and output.We trained LSTM network,CNN network and LSTM-CNN network with the same experimental data and number of layers.The results have proven that LSTM-CNN network exceeds in training speed and fitting accuracy.

关 键 词：深度学习 同步电机 励磁绕组匝间短路 故障预警 LSTM-CNN网络 

分 类 号：TM341[电气工程—电机]